Design and measurements of one ring array

In a linear array, it was shown that using a single cavity for the whole array is more convenient (Guinvarc’h et al., 2012) than using individual cavities, as presented in Fig. 1.18 (pg. 16).

The dimensions and configuration of one ring array with its cavity are shown in Fig. 3.13. Only spirals of one polarization are depicted to simplify the scheme. In fact, there are 8 antennas. 4 antennas of right hand circular polarization were interleaved with the other 4 antennas of left hand circular polarization, in order to obtain dual polarization. The antennas of right hand circular polarization are fed using the sequential rotation technique.

An Archimedean spiral over FR4 substrate of 0.81 mm of thickness was fabricated in coordi- nation with PhD Karim Louertani from NUS Temasek Lab, Singapore. The substrate is used to reduce the input impedance of the spiral. Hence, the antennas of left hand circular polarization are terminated with 100Ω. As in section 1.3.3 (pg. 14), all the antennas have a diameter of 10.5

Figure 3.13: Scheme and dimensions of one ring spiral array. Only spirals of one polarization (blue) are depicted to simplify the scheme.

Figure 3.14: Simulation of reflection coefficients of a single Archimedean spiral antenna. In blue with no cavity, in green when using small cavity (cf. Fig. 1.18, pg. 16) and in red when using the large cavity (cf. Fig. 3.13). (Zref = 100Ω)

reduce the coupling between the border of the cavity and the spiral. The distance between the bottom of the cavity and the spirals was also kept at 5 cm. The distance between the spirals of the same polarization is 21.92 cm.

Effect of the array cavity on the reflection coefficient

Fig. 3.14 depicts the reflection coefficient of a single spiral with different cavities to compare their effect. We observe that the presence of the cavity, small or large, slightly improves the

Figure 3.15: Simulations of reflection coefficient of one ring array, with and without large cavity (blue and green, respectively), and a single spiral with large cavity (red). (Zref = 100Ω)

lower frequency limit of the reflection coefficient, being better when using the small cavity.

Fig. 3.15 shows the simulation results of one ring array of Fig. 3.13, with and without cavity, and compares it with the single spiral backed by the cavity of the array. In the case of the array, the improvement in the reflection coefficient is due to the presence of the cavity and the coupling with the other spirals. But the array cavity also affects the input impedance of the spiral increasing the reflection coefficient slightly over -10 dB at 1.35 GHz and 2.1 GHz. This issue can be overcome by optimizing the matching of the element in the cavity and the feeding system.

Figure 3.16: Prototype of the one ring array (cf. Fig. 3.13). Right hand spirals are marked with red color, left hand spirals are marked with blue color.

Prototype

As depicted in Fig. 3.16, the Archimedean spirals were placed over a foam of 5 cm height (Hspiral). To create the cavity, two strips of copper, with height of 3 cm (Hwall), were put around the spiral ring with inner radius of 8.7 cm (Rin), outer radius (Rout) of 22.3 cm. The left hand polarization spirals were terminated with chip resistors of 100Ω. The feeding system of section 1.3.2 (pg. 11) was used.

Figure 3.17: Simulation and measurements of ref. coef. of the one ring array of Fig. 3.16 (Zref = 100Ω).

Figure 3.18: Total gain (simulation and measurements with 2.5dB of losses correction) of the one ring array at broadside of Fig. 3.16.

Measurements

Fig. 3.17 compares the reflection coefficient of the measurements and simulations. The mea- surements also show the mismatch at 1.35 GHz. Since the diameter of the symmetrical spiral antenna is 10.5 cm, using Eq. 1.11 (pg. 11) along with the p factor of the Archimedean spiral

antenna with no substrate (pS11 = 1.15, cf. Tab. 1.3, pg. 11), we expect itsS11 cutoff frequency at 1.05 GHz.

(a) φ= 0^◦1000 MHz (b) φ= 0^◦ 1200 MHz

(c) φ= 0^◦ 1400 MHz (d) φ= 0^◦ 1600 MHz

(e) φ= 0^◦ 1800 MHz (f) φ= 0^◦2000 MHz

Figure 3.19: Cuts of radiation pattern at φ = 0^◦ of Fig. 3.16. Simulation results with FEKO in blue. Measurements in red.

(a) φ= 45^◦ 1000 MHz (b) φ= 45^◦ 1200 MHz

(c) φ= 45^◦ 1400 MHz (d) φ= 45^◦ 1600 MHz

(e) φ= 45^◦ 1800 MHz (f) φ= 45^◦ 2000 MHz

Figure 3.20: Cuts of radiation pattern at φ = 45^◦ of Fig. 3.16. Simulation results with FEKO in blue. Measurements in red.

In fact, we can see that the reflection coefficient of the spiral with substrate starts a bit earlier. The S11 bandwidth goes from 0.96 up to 2.1 GHz. We observe a good agreement of the experimental results with the simulation.

The radiation patterns of the simulations and measurements can be seen in Fig. 3.19 (pg.

89) and 3.20 (pg. 90). The measurements are presented after compensation for the feeding system (about 2.5 dB). The curves agree with some differences in the sidelobes. The total gain in the simulations is about 12 dB (cf. Fig. 3.18). This is expected since the gain of a single Archimedean spiral in free space is about 4 dB (cf. Fig. 1.10, pg. 10), using 4 spirals gives 6 dB more and the cavity gives, at most, 3 dB more.

Figure 3.21: XpolR at broadside of one ring array of Fig. 3.16.

Figure 3.22: Relative side lobe level of the one ring array of Fig. 3.16.

The XpolR at broadside (cf. Fig. 3.21) is very good (XpolR > 15 dB) between 1 GHz and 2 GHz, as expected. The difference between the simulation and the measurements is due to the errors in phase and amplitude introduced by the hybrid couplers and power dividers used in the

feeding system. Even if the reflection coefficient at 1.35 GHz presents a mismatch, the XpolR is still good thanks to the sequential rotation technique.

Fig. 3.22 shows the relative side lobe level (RSLL) of the ring array. Since the distance between the elements is 21.92 cm and there is no scanning, it is expected to have grating lobes at 1.37 GHz, for a square lattice infinite uniform array. Since in this array there are just 4 antennas per polarization the grating lobes appear earlier. The measurements show that the side lobes reach the limit of -10 dB at 1.1 GHz. Hence, the RSLL bandwidth goes from 1 GHz to 1.1 GHz.

In the next section we will see how using more rings reduces the side lobes level to less than -10 dB, relative to the main lobe. In this case, the useful bandwidth is strongly dependent of the RSLL (cf. Fig. 3.23).

Figure 3.23: Summary of the bandwidths of the one ring array.

The measurements and fabrication of this array were performed in coordination with PhD Karim Louertani, from NUS Temasek Laboratories, Singapore.